At its latest announcement, the supply chain visibility giant introduced Loft, an AI-native orchestration platform designed to work across any enterprise system—not just supply chain stacks. At the center of Loft is Sophie, an AI “developer agent” that translates natural language operational requirements into production-ready automations in days, rather than the months traditional deployments demand.

If that promise holds, it tackles one of enterprise AI’s most stubborn realities: scaling beyond pilot purgatory.

The Scaling Crisis in Enterprise AI

Enterprise AI adoption is widespread—but shallow.

McKinsey reports that 88% of organizations have deployed AI somewhere, yet only 7% have scaled it enterprise-wide. Gartner predicts 40% of agentic AI projects will be abandoned by 2027 due to complexity and unclear ROI. Deloitte adds that 70% of enterprises take more than a year to resolve post-deployment AI maintenance challenges.

The pattern is familiar: organizations deploy AI agents layered atop fragmented systems—ERP here, TMS there, Slack threads everywhere. AI can observe problems, but acting across systems requires brittle integrations, engineering resources, and constant oversight.

Josh Jewett, operating partner at NewRoad Capital Partners and former CIO of Dollar Tree and Family Dollar, described the issue succinctly: critical decision logic often lives outside systems entirely—buried in spreadsheets, inboxes, and chat threads. When AI is layered on top of that fragmentation, it struggles to act with authority.

Loft is FourKites’ answer to that structural challenge.

From AI Features to AI-Native Orchestration

Unlike point AI features embedded inside existing applications, Loft is positioned as an AI-native orchestration layer.

It works across ERP, ITSM, TMS, WMS, and CRM systems, while simultaneously pulling in real-time external intelligence from the FourKites Intelligent Network—which includes insights from over 500,000 trading partners across 176 countries and processes approximately three million supply chain events daily.

That external data layer is FourKites’ core differentiator.

Most enterprise AI agents operate solely on internal enterprise data. Loft combines internal system orchestration with real-time network intelligence—supplier performance, carrier reliability, manufacturing disruptions, and capacity constraints that no single enterprise system contains.

As Charles Brennan, Senior Analyst at Nucleus Research, notes, the value of automation depends on the data foundation behind it. FourKites’ network provides context beyond the four walls of the enterprise.

Meet Sophie: The AI Developer Agent

At the center of Loft is Sophie, designed to function as an AI developer agent.

Here’s how it works:

• Customers describe operational requirements in natural language.

• Sophie determines whether existing workflows can be configured.

• If needed, she combines reusable building blocks or recommends custom code.

• FourKites engineers review before deployment.

• Sophie continues monitoring performance post-launch.

Instead of months-long engineering cycles, automations can move from idea to deployment in days. Just as important, Sophie continuously improves workflows over time—addressing model drift and performance degradation that typically create ongoing engineering tax.

That “maintenance elimination” pitch is key. Many enterprises discover that the real cost of AI comes after go-live.

Agent Operating Procedures: Capturing Decision Logic

Loft introduces a concept called Agent Operating Procedures (AOPs).

When AI agents handle tasks—resolving purchase order mismatches, escalating supplier delays, balancing warehouse capacity—the platform records not just what decision was made, but why. It captures context, precedent cases, and human approvals.

In most enterprises, that reasoning disappears into chat threads and email chains. Loft aims to preserve it as structured, reusable logic.

The result is cumulative intelligence: each decision makes the next one easier.

The Digital Workforce in Action

Loft also houses FourKites’ existing “Digital Workforce,” including specialized agents like:

• Tracy for logistics execution

• Sam for supplier collaboration

• Alan for appointment scheduling

These agents are already deployed at dozens of Fortune 500 companies, according to FourKites. Sophie expands the framework by enabling rapid creation of new automations tailored to specific operational requirements.

Under the hood, Loft is built on the same architecture as the FourKites Intelligent Control Tower, combining:

• Network data

• Digital twins

• A digital workforce

The platform pulls data from more than 200 enterprise systems to power cross-functional automations that respond dynamically to real-world conditions.

Why External Intelligence Is the Moat

The broader AI agent market is becoming commoditized. Foundational models are widely accessible, and vendors increasingly rely on similar infrastructure stacks.

FourKites’ bet is that durable differentiation lies in proprietary data access—specifically, external supply chain intelligence at scale.

When an AI agent decides whether to escalate a supplier delay, Loft doesn’t rely solely on internal metrics. It factors in that supplier’s real-time performance across the network, patterns from other customers, and historical precedents.

That external reality layer turns AI from reactive analytics into predictive, cross-enterprise orchestration.

From Dashboards to Autonomous Execution

FourKites CEO Mathew Elenjickal frames the shift clearly: enterprises must move from dashboards that track problems to systems that autonomously solve them.

Loft represents an attempt to close the gap between AI insight and AI action—while reducing the engineering burden that often derails scaling efforts.

If successful, it could push enterprise AI from experimentation to durable operational infrastructure.

And in a market where many agentic AI initiatives may stall by 2027, durable may be the operative word.

Get in touch with our MarTech Experts.